Reproduce “Accelerating Video Diffusion via Training-Free Trajectory Routing” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Mustafa Munir, Huy Vu, Shreyas Misra, Rohit Jena, Sajad Norouzi, Ali Taghibakhshi, Anis Ahmad, Anjul Patney and 2 more
TRACK routes denoising steps in video diffusion between compatible large and small models. During offline calibration, it compares their predictions on shared latents and assigns the small model to steps with least disagreement, then follows that schedule without changing step count. On Wan 2.1, it routes 30 of 50 steps small, achieving 1.95× speedup with comparable four VBench scores, subject to checkpoint compatibility and higher memory use.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.